Thread by @mbeisen

What I learned from years on graduate admissions committees is that they don’t predict success - they determine it. Everyone has a pet theory, rarely based on evidence, and never based on good evidence, about what makes a successful student.

And, because the admitted pool is enriched for students who meet whatever criteria happen to be in the ascendancy, and because some of these students succeed, we convince ourselves that we were right and keep doing it.

I’m not saying that everyone is equally likely to succeed in graduate school in its current form and that there are not predictors of success. I am saying we don’t - and given our methods can’t - know with confidence what they are.

And I don’t claim to be free of this - I have things I look for (people who succeeded in situations where success was not handed to them) - but that’s predicated on liking to work with such students and wanting to give them a chance as well as on belief they’ll do well.

Which brings me to a hypothesis that I wish I had data to interrogate - that we as individual scientists are better at identifying students who will do well as our individual trainees than we are at identifying students who will do well in a generic lab.

And thus I’ve grown increasingly uncomfortable with the working of admissions to the large umbrella programs that dominate molecular biology.

I see the advantages of organizing graduate training this way for the students - but in my experience the admission process tends towards a kind of consensus based uniformity in the types of students who get opportunities that is counterproductive for students and science.

And I think that homogenization is a product of people thinking they know how to identify people who will succeed “in graduate school” as opposed to in their labs.

Hence I am extremely leery of any effort anywhere to try to query and hence build broader consensus on “what predicts success in graduate school”, especially when it’s based, at best, on anecdata.